Catalog

Detect and fix stale chunks (files that changed or were deleted since last indexing)

Autonomous error recovery — detect failures, diagnose root cause, apply fixes, and resume without stopping

AI coding agents fail in consistent, predictable ways. They fabricate npm packages that don't exist. They catch errors and continue silently. They add features you never asked for. They lose context mid-session and forget what they were doing.

Multiple AI agent sessions working on the same repository cause chaos. Session A is on branch feature/auth. Session B switches to main to check something. Session A's next tool call fails because the branch disappeared. Or worse: Session B force-pushes over Session A's work.

Composite skill — full project health check across testing, config, hooks, performance, security, MCP, and plugins. Runs the audit skills in parallel and reconciles into one severity-ranked report with prioritized remediation plan. Use weekly per active project, before major releases, or as part of quarterly tech-debt review.

Git-level WIP safety net. Stashes uncommitted work, tags, and pushes to remote

Composite skill — full debugging workflow from "this is broken" to root cause and fix. Chains systematic-debugging (root-cause hypotheses) → tracer agent (evidence walk) → sentry (production correlation if applicable) → ci-watch (regression check) → incident-response (if production-impacting). Use when a bug needs deep investigation, not just a quick fix.

End-to-end performance audit composite — baseline the slow metric, profile to find the real hotspot, implement the smallest targeted fix, re-benchmark to confirm improvement, capture in ADR. Use when users report slowness, CI shows runtime regression, or before a release. Never optimize without profiling first.

Manually improve corpus quality by adding missing docs and filling retrieval gaps

Verify work before committing, pushing, or requesting review

Recover a PR whose head SHA disagrees with its branch ref (webhook desync). Force-push and close+reopen rarely fix it — recreate the PR.

Composite skill — one-pass cleanup of stale local branches, dead worktrees, merged branches, and abandoned remote PR branches. Chains `git fetch --prune` → `clean_gone` (kill [gone] branches) → worktree prune + offer-to-remove dead worktrees → list-and-delete branches merged to main and release → delete remote PR branches whose PRs merged >7 days ago. Use instead of running `clean_gone` alone — that only catches half the rot. Daily-friction composite; fires on "clean up branches", "branch hygiene", "stale worktrees", and on session start when local branch count > 30.

Composite skill — configure a fresh or under-configured repo for the release-branch workflow in one pass. Creates the `release` branch on origin, seeds `CHANGELOG.md` (Keep-a-Changelog format), drops in `.claude/dep-sweep-config.json` + `.claude/branch-hygiene-config.json` + a `release-cadence.md` standard stub, opens a `chore/repo-bootstrap` PR to main. Use instead of manually setting these up before the first `/pr-to-release` — without this, none of the release-branch composites can fire. Idempotent: skips any artifact that already exists. Replaces ad-hoc "set up the new repo" scripts.

Meta-skill — defines when to apply other skills automatically without being asked

Orchestrate teams of agents — DAG execution, routing, state sharing, failure recovery. Use when coordinating parallel agents in task pipelines with dependencies or distributing work by capability.

Coordinate multiple subagents/worktrees for parallel workstreams. Decomposes larger tasks into independent sub-tasks, dispatches each to a dedicated agent.

Examine what's actually stored in the index for specific items

Evaluate retrieval quality from the local RAG index

Onboard into an unfamiliar repository fast before making any changes

Audit historical Claude Code token usage from JSONL session files — shows total spend, cache hit rates, top costly sessions, weekly trends, and where tokens are going. Run after heavy sessions or weekly to measure optimization impact.

Design and author reusable subagents for specialized AI tasks. Use when the task needs an isolated agent persona, scoped system prompt, or reusable domain-specific assistant rather than a general skill.

Execute phased plans with multiple independent tasks per phase by fanning out one agent per task, reconciling outcomes per wave, gating between phases with verify commands, and emitting a phase × outcome report. Triggers include "execute this plan", "work through these phases", "swarm over this backlog", "parallelize this plan".

Composite skill — design, scaffold, build, and verify a UI in one workflow. Chains ui-ux-pro-max or web-design-guidelines (audit/principles) → shadcn or tailwind-design-system (component scaffold) → impeccable or frontend-design (build) → webapp-testing (verify). Use when the task is "build this page", "design and implement X", or any new UI surface.

Diff the Anthropic `claude-plugins-official` marketplace against installed plugins and surface the delta, filtering known-unwanted entries (language LSPs you don't use, SaaS integrations you don't use, sample/demo plugins).

Audit corpus distribution by source type and repo; identify coverage gaps and underindexed topics

Ship validated changes with repeatable release evidence — version bump, changelog, tag, and optional GitHub release

Audit and prune a bloated test suite down to the minimum tests that hit the coverage threshold and guard real behavior. Replaces many shallow unit tests with fewer well-scoped integration tests. Use when test count is disproportionate to app size.

AI agents are only as good as the context they receive. A powerful model with zero project context produces generic code. A mediocre model with excellent context produces code that fits your project perfectly.

Use when creating hooks for Claude Code plugins. Covers hook events, configuration, exit codes, and stdin/stdout contracts. Hooks execute in response to events to validate operations and enforce policies.

Full dependency lifecycle composite — audit known vulnerabilities, triage by severity and breaking-change risk, upgrade targeted packages, verify tests pass, ship PR. Use monthly, before releases, or when dependency alerts are stacking. Never do a blanket upgrade; always triage first.

Composite skill — diagnose, repair, and validate a test suite end-to-end. Chains test-health (diagnose) → config-drift-detect (gate compatibility) → test-cleanup (prune + add integration tests) → mutation-test (validate survivors) → adr-write (capture decisions) → docs-sync. Use when "the test suite is bad" or you've hit the test-cleanup-bails-at-the-gate failure mode.

Run mutation testing against a project's test suite to verify the tests actually catch failures when source code is broken. Use after major test changes, before declaring a suite "good", or when coverage % looks healthy but you suspect tests aren't catching real bugs.

Decide the highest-value safe thing to do right now in the active repo or workspace.

Sync durable project or session knowledge into the available memory systems so future sessions have accurate context. Use when meaningful work is complete and the user wants the result remembered.

Full test suite improvement composite — audit health, prune dead tests, validate with mutation testing, capture rationale in ADR. Use after a major feature ship, before tightening coverage gates, or when the suite shows bloat, excessive skips, or slow runtime. Chains test-health → test-cleanup → mutation-test → adr-write.

This skill should be used when the user asks to "create an agent", "add an agent", "write a subagent", "agent frontmatter", "when to use description", "agent examples", "agent tools", "agent colors", "autonomous agent", or needs guidance on agent structure, system prompts, triggering conditions, or agent development best practices for Claude Code plugins.

Composite skill — runs the postmortem chain after any production incident (`/hotfix`, rollback, or prod outage acknowledged). Chains adt-research (root-cause learning) → adr-write (decision capture) → generate-tests (regression test) → security-sweep (conditional, only if root cause is auth/input/secret-related) → knowledge-loop (memory + RAG curation) → handoff. Stops the silent-postmortem failure mode where a hotfix ships and the lessons evaporate. Auto-queues after `/hotfix` Phase 10 completes; also fires when user says "postmortem", "what did we learn", "write up the incident".

SKILL.md is an open specification released by Anthropic in 2025 that standardizes how tools discover and auto-invoke procedural knowledge (skills, patterns, guides, rules). Instead of each tool (Claude, Cursor, VSCode, CLI) maintaining isolated skill registries, SKILL.md enables a single skill to be published once and auto-invoked across any tool that speaks the spec.

Composite skill — query, capture, improve, and persist knowledge in one workflow. Chains recall (RAG query) → sync-memories (write durable note) → rag-curate (improve weak retrievals) → handoff (durable snapshot if session-ending). Use when the work involves "what did we decide", "remember this", "save where we are", or any closing checkpoint.

Composite skill — full security pass across secrets, dependencies, code paths, and OWASP risks. Chains security-audit (broad) + socket-audit (npm supply chain) + semgrep (pattern scan) + code-security (code review for vulns) in parallel, reconciles into one severity-ranked report with remediation plan. Use quarterly per active repo or before any release of security-sensitive code.

Composite skill — full SEO + Accessibility audit of a website or webapp in one workflow. Chains audit-website (squirrelscan SEO + tech + perf scan) + seo-audit (SEO framing) + playwright-best-practices (axe-core a11y) + ui-audit (design-system accessibility patterns) → reconciled severity-ranked report split by SEO vs A11y vs Both. Use for periodic site health checks, pre-launch audits, or when search rankings drop.

Decision guide for when to proactively use Claude Code slash commands — /think, /model, /compact, /clear. Use when choosing whether to invoke a command before or during a task, or when a complexity-classifier hook has flagged a task as high or critical. Pairs with the complexity-classifier UserPromptSubmit hook.

Sync files (skills, packages/core/patterns, agents, specs) from EN canonical repo to PT mirror, inserting "Tradução pendente" blockquote after YAML frontmatter, then open PR.

Audit Claude Code hooks for fire frequency, latency, exit codes, and output value. Surfaces hooks that never fire, hooks that fire too often (spam), hooks that exit non-zero silently, and hooks that fire but produce no output Claude actually uses. Run weekly automatically. Required after wiring new hooks.

You have 5 projects, each with a session. You open each one, paste "continue with next priorities", wait, switch, repeat. You're acting as a human cron job.

Losslessly compress PNG and SVG assets in place and generate .webp siblings for every raster image. Use when the user asks to "compress assets", "optimize images", "shrink the assets folder", or "convert to webp". Runs a deterministic pipeline via oxipng + svgo + cwebp + ImageMagick.

Keep sessions focused and efficient — prune stale outputs, preserve active state, recommend compaction

React composition and performance optimization patterns from Vercel. Use when building component libraries, refactoring components with prop proliferation, optimizing React/Next.js performance, or designing scalable component architectures. Covers composition patterns, React 19 APIs, performance optimization across waterfalls, bundle size, server/client rendering, and re-render optimization.

Use when creating, writing, or organizing slash commands. Covers frontmatter, dynamic arguments, file references, and bash execution. Commands are instructions FOR Claude, not messages TO the user.

Composite skill — first-touch workflow for an unfamiliar repo. Chains adt-repo-intake (understand) → audit-deep (health check) → config-drift-detect (gate audit) → init (CLAUDE.md if missing) → fewer-permission-prompts (prune permission noise). Use when entering a new repo for the first time, taking over from another developer, or after a major repo restructure.

How you ask the model matters more than which model you ask. Two disciplines stop most AI coding failures before they start.

Mega-composite — full greenfield feature development from idea to live. Chains research-and-decide (validate the idea + capture decision) → scope-and-execute (analysis + plan + implement) → design-build (if UI involved) → fix-the-suite (if tests need work) → merge-confidently → ship-it → knowledge-loop. The single command for "build feature X end-to-end."

Composite skill — take a PR from "I think it's ready" to merged with full gate verification. Chains pr-merge-readiness (verdict) → ci-watch / gh-fix-ci (if blockers) → gh-address-comments (if review needed) → ship (when verdict is MERGE). Use as the one-call answer to "can I merge this?" instead of running 4 skills sequentially.

Trigger a full or incremental reindex of the RAG corpus

Composite skill — safely refactor a module end-to-end with sequencing, parallel implementation, post-refactor cleanup, and rationale capture. Chains refactor-plan (phased plan + rollback) → three-man-team (architect/builder/reviewer in parallel) → fix-the-suite post-refactor → adr-write → docs-sync. Use for non-trivial refactors that need both careful sequencing and durable record.